16:15
Midjourney prompts
Olivio Sarikas presents a curated collection of his best Midjourney V6 and Niji 6 prompts — covering photoreal pets, glass figurines, anthropomorphized dog portraits, romantic landscapes, comic-hybrid characters, Disney-style baby animals, isometric miniature worlds, Valentine's couples, noir motion shots, fisheye action shots, and dreamy surreal scenes — and shows that most of the same prompts, with some adaptation, also produce strong results in Stable Diffusion SDXL.
Photo-realistic hamster prompt: 'photo realistic cute hamster wearing colorful hawaiian shirt and expensive golden sunglasses holding a glass of champagne in one paw standing on an expansive teak wooden floor out on the ocean, great detail, extra wide angle, hd, 8k' — works in both Midjourney and SDXL turbo, though Midjourney's champagne often reads more like beer.
Glass figurine prompt: 'super cute glass figurines of a cute animal' — cute cuddly pets (seals, lions, dolphins) work best; Midjourney gives varied randomized styles, SD gives a beautiful figurine with a less lively background that can be added back into the prompt.
Anthropomorphized dog portrait prompt: 'highly stylized and anthropomorphized portrait of a [dog breed]... humanlike qualities positioned in a regal pose against a floral background... wearing round ornate sunglasses... richly embroidered floral attire... noble aristocratic portrait playfully reimagined' — any dog breed can be swapped in; the SD version reads more like a painting than a photo.
1960s Space Age dog-fashion prompt: front-view portrait of a dog breed in 1960s mid-century Space Age fashion, deep blue backdrop, matching-color lounge coat and sunglasses — flagged as good for memes and social posts; the SD version's clothing looks 'a little bit more plasticky.'
Romantic landscape/getaway prompt: 'a beautiful romantic [church/house/lake]... flowers, distant view, 8K dreamy digital art, cozy style with warm colors' (contains an on-screen typo, 'style', that Olivio notes still works) — keeping flowers in the foreground adds depth; reused across autumn, winter, and farmhouse scenes.
Comic-book line-art human/animal hybrid prompt: 'comic book color sketch line art style, an adorable human hybrid [animal] man holding a walking stick' — the stick can swap for a sword; tried with a lion, frog, snake, and toad; the SD version needed extra wording ('wearing Han Fu', 'standing') to hit the intended look.
Disney/Pixar-style baby-animal render prompt: 'cute Disney design style render of an animal baby [subject] with a round body and large glowing pupils, edge light, soft pastel background' — Olivio warns not to push the modifications too far or the effect breaks; the SD version is less 'cute' since it lacks the Disney/Pixar render style.
Isometric miniature-world multi-prompt (e.g. 'old lady standing in an isometric mini world pop out effect in 3D') using Midjourney's double-colon (::) syntax to split the prompt into weighted parts — flagged as unstable/non-reproducible; one variant became one of Olivio's viral tweets.
Valentine's Day embracing-couple prompt: 'pixar style animation, [subject] and [subject] are embracing each other' — subjects, genders, and ethnicities can be swapped to cover many combinations; the SD version needed tweaks like 'happy old men'/'happy old woman' to fix odd facial expressions and still produced a hand with only four fingers.
Graphic-novel noir motion prompt, rendered with Niji 6 rather than V6: 'graphic novel noir style illustration of a [subject], fast movement, high-speed marker sketch illustration, black white and red color palette... motion trails, low angle perspective, neutral black background' — applied to a dancer, a raging lion, and a musician; also renders well in SD.
Fisheye dynamic-action prompt: 'dynamic power movement, fast speed, cinematic dramatic super fisheye shot of a [subject]' — used for a skateboarder, a racing car, a motorcyclist riding into a sunset, and a superhero (the Flash); the SD version looks 'a little bit different' and could use more color.
Extreme absurd fantastic fisheye close-up prompt: 'very very extreme absurd fantastic [subject], extreme closeup with fish eye, dramatic camera angle, dramatic blue light neon effect, colorful fashion shot, highly detailed' — used for a dragon/lizard woman and similar creature close-ups; Olivio says he 'could not 100% replicate' this one with Stable Diffusion.
Dreamy surreal-scene prompt: 'very extreme strange good dream scene, colorful' — described as erratic/random but visually striking, dreamlike, with floating bubble elements; adding 'landscape' pushes it further; also produces strong results in SDXL.
Distribution note: the full prompt text for every image is posted free on Olivio's Twitter/X account; Patreon supporters additionally get a downloadable PDF collecting all the prompts.
The SD comparisons consistently show a turbo SDXL model giving a faster, lower-detail baseline, with quality improving when rendered on a full (non-turbo) SDXL model at more steps and then upscaled.
Several prompts do not transfer verbatim between Midjourney and SD — Olivio had to manually add words ('standing', 'wearing Han Fu', 'Happy old men'/'Happy old woman') to get comparable SD results.
The double-colon (::) multi-prompt syntax lets one Midjourney prompt be split into separately weighted segments, but Olivio explicitly flags it as unstable and not reliably reproducible.
Switching from Midjourney V6 to Niji 6 (not just changing prompt wording) is what produces the graphic-novel/marker-sketch look in the noir motion prompt — model choice is treated as its own lever alongside prompt text.
The SD render of the Valentine's couple prompt has a visible anatomical flaw (four fingers on one hand), which Olivio flags as needing inpainting or further correction.
Olivio warns that pushing a working prompt with too many additional modifications can break the intended effect, citing the Disney-style baby-animal prompt as the example where he advises restraint.
«I have a collection of amazing prompts for you»
— 00:00
«my patreon supporters get a PDF with the collection of these prompts but if you want to have them for free you can follow me on Twitter»
— 00:12
«this one is not a very stable prompt this is why here I have put the prompt every single time»
— 08:22
«here you have these double colons this is creating a multi- prompt so this is kind of breaking up the prompt into multiple parts of a prompt and it works a little bit better but it's not very stable»
— 08:39
«this is actually one of my tweets that went viral»
— 09:10
«don't go too far or it might not work anymore as the image as intended»
— 08:48
«this is something that really took me by surprise very interesting something I could not 100% replicate with stable diffusion»
— 13:48
«it was an absolute blast to experiment with these prompts for you»
— 15:49
Reception
Audience appreciates the video's quality and prompts with engaged discussion about AI tools, though some debate the relative merits of Midjourney versus Stable Diffusion.
The video delivers a straightforward, well-organized prompt-recipe roundup with concrete copy-paste text and side-by-side Midjourney/SD comparisons for each image, consistently framing the Stable Diffusion results as secondary and requiring adaptation relative to the Midjourney originals.
16:15